Russian internet company Mail has launched an AI-powered support chatbot that now resolves 74% of user queries without involving human operators. The chatbot leverages Retrieval-Augmented Generation (RAG) technology, achieving a 77% accuracy rate for its responses.
- Since launch, the volume of queries handled by the chatbot has increased 6.5 times.
- The chatbot handles 31% of all support requests, becoming the most popular contact point for users.
- The accurate and helpful response rate stands at 77%.
The chatbot integrates large language models with Mail’s extensive help center database using RAG technology, which combines AI-generated content retrieval with natural language generation. This approach classifies requests against frequent user scenarios and help articles, allowing the chatbot to provide instant answers and practical recommendations.
In parallel, machine learning models collect necessary information and open tickets for complex cases, ensuring human operators receive detailed context when stepping in. This drastically cuts down repetitive questioning and accelerates problem resolution.
Pavel Andreev, head of user support for Mail and Mail Cloud, said, ”Since launching the chatbot, the number of requests it handles has grown 6.5 times. Now, 74% of user queries are resolved without operator involvement, and 77% of responses are accurate and helpful. Thanks to RAG and machine learning, users get faster solutions independently, while our specialists focus on more complex issues.”
Mail’s move mirrors a growing trend among major tech companies integrating AI chatbots into customer support to reduce wait times and improve efficiency. Compared to global players like Google and Microsoft, who also employ RAG-based models, Mail’s chatbot stands out in the Russian market for its rapid adoption and high automation rate.
As AI chatbots continue to advance, the challenge will be balancing automation with quality human support. Mail’s next steps will likely focus on refining the chatbot’s accuracy beyond 77%, handling increasingly complex requests, and expanding its role across more services.